programmatic-seo · fcakyon — Claude Code skill: what it does, how to install
programmatic-seo plans scalable template-driven page generation from data sets and protects your project from thin-content deindexing penalties.
MIT ★ 1,216 updated 2026-11-09
/plugin marketplace add fcakyon/claude-codex-settings
/plugin install seo-skills@claude-settingsWhat programmatic-seo does
This skill guides the rollout of programmatic pages built from templates and structured data. It assesses data defensibility across five tiers, prioritizing proprietary telemetry over public records. You also get criteria to pick appropriate patterns, such as directories or comparisons.
To keep search crawlers satisfied, the guidelines advise branching templates instead of padding sparse pages. Navigation relies on a hub-and-spoke model in subdirectories rather than subdomains. Linking rules prevent orphan items by ensuring deterministic catalog fallbacks.
Publishing thousands of variations at once risks spam flags. Wasteful URLs squander crawl budget. The workflow requires deploying high-demand slices first with dedicated sitemaps to verify indexing rates before scaling out.
What it helps with
- select the right page layout pattern for a specific dataset
- prevent doorway page penalties and search engine deindexing
- organize hub-and-spoke internal links without orphan pages
- stagger batch publication to track indexation rates
Example prompts for the agent
- Review our integration dataset and pick a pattern for category hubs
- Audit this directory template to make sure thin rows do not break indexation
- Draft an internal linking blueprint for our location-based service pages
How to install programmatic-seo
1Plugin marketplace
/plugin marketplace add fcakyon/claude-codex-settings/plugin install seo-skills@claude-settingsThis installs the whole seo-skills plugin with all its skills
2Skills CLI (Vercel)
npx skills add fcakyon/claude-codex-settings --skill programmatic-seo -a claude-code3Manually: put the files in place
~/.claude/skills/programmatic-seo/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/programmatic-seo/
1Skills CLI (Vercel)
npx skills add fcakyon/claude-codex-settings --skill programmatic-seo -a codex2Manually: put the files in place
.agents/skills/programmatic-seo/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add fcakyon/claude-codex-settings --skill programmatic-seo -a cursor2Manually: put the files in place
.cursor/skills/programmatic-seo/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install fcakyon/claude-codex-settings plugins/seo-skills/skills/programmatic-seo2Manually: put the files in place
.github/skills/programmatic-seo/SKILL.md1Skills CLI (Vercel)
npx skills add fcakyon/claude-codex-settings --skill programmatic-seo -a gemini-cli2Platform CLI
gemini skills install https://github.com/fcakyon/claude-codex-settings --path plugins/seo-skills/skills/programmatic-seoManually: put the files in place
.devin/skills/programmatic-seo/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about programmatic-seo
How to install programmatic-seo in Claude Code?
Run /plugin marketplace add fcakyon/claude-codex-settings, followed by /plugin install seo-skills@claude-settings.
What does programmatic-seo do?
It gives Claude Code guidelines to architect scalable data-driven pages, verify search intent, and avoid thin-content crawl traps.
Why does the skill recommend subfolders instead of subdomains?
Subfolders consolidate search authority directly on the parent domain, whereas separate subdomains dilute topical weight.
Source text
SKILL.md · MIT
--- name: programmatic-seo description: "Use when planning many similar pages from a template and a data set: choosing the pattern, deciding whether the data can carry it, and avoiding the thin-content failure that gets page sets deindexed." license: MIT --- # Programmatic SEO Building pages from a template and a data set works when each page answers a real query with something only your data can say. It fails the same way every time: a template with the variable swapped, published at volume, on queries nobody searches. That is the definition of a doorway page, and Google's spam policies name it. So the question to settle before any of the mechanics is whether the data is worth a page. ## Does the data carry it Defensibility runs in this order, strongest first: 1. Proprietary, because you generated it. 2. Product-derived, from how your own users behave. 3. User-generated, from your community. 4. Licensed, where the license is exclusive. 5. Public, which anyone can use and everyone already has. A set built on public data competes with every other set built on the same public data, so it has to win on presentation, freshness, or aggregation instead. That is possible and it is a much harder brief. Say so rather than shipping 5,000 pages that restate a public API. Then check demand actually exists: aggregate volume across the pattern, how it splits between head and long tail, and whether the trend is going anywhere. A pattern with 10,000 combinations and no searches for 9,000 of them is a 1,000-page opportunity. ## Choosing a pattern | You have | Consider | | --- | --- | | Proprietary data set | Directory, Profiles | | A product with integrations | Integrations | | A design or creative product | Templates, Examples | | Several distinct audience segments | Personas | | A local footprint | Locations | | A utility or calculator | Conversions | | Deep subject expertise | Glossary, Curation | | A crowded competitive field | Comparisons | `references/playbooks.md` covers the twelve patterns, each with its query shape, what data it needs, and how it usually fails. Patterns layer, and layering is often where the real opportunity is: "best coworking spaces in Austin" is Curation crossed with Locations. ## Making each page worth indexing - Every page needs something specific to it beyond the substituted variable: a number, a comparison, a list, an image, a genuine answer. - Write the introduction per page, or generate it from enough fields that no two read alike. A shared paragraph with one word swapped is the tell. - Vary structure by what the data supports. A page with three data points shouldn't use the layout built for thirty; branch the template instead of padding. - Give each page a unique title and meta description built from its variables, not one pattern with the variable appended. - Match the query's intent. A comparison query wants a comparison, not a signup page. ## Structure and indexation - Subfolders, not subdomains, so the pages build authority on the same domain rather than splitting it. - Hub and spoke: a category page linking to every page in the set, each page linking back, and related pages cross-linking. Otherwise the set is orphaned and only the sitemap points at it. - Own sitemap for the set, or a sitemap per pattern, so indexing rates per pattern are readable. - Ship the strongest slice first rather than the whole set. Publish the highest-demand pages, confirm they get indexed and ranked, then expand. A set that goes out all at once gives no signal about which part worked. - Leave the thinnest variations out entirely. That is better than publishing them and noindexing them later, since crawl spent on pages you didn't want is crawl not spent on pages you did. ## Before it ships Check that each page has unique value and answers its query, that titles and descriptions are unique, that heading structure and schema are in place, that the set is linked from the site rather than only the sitemap, that no page has a conflicting noindex, and that nothing in the set competes with an existing page for the same query. After launch, watch what fraction of the set gets indexed, which patterns rank, and whether engagement holds up. Thin-content warnings, a drop across the set, or an indexing rate that stalls well below 100% all mean the same thing: cut, don't add. ## What this hands off Programmatic SEO decides the pattern, the data, the URL shape, the template's sections, and the linking. The words inside a template section are copy: hand the content marketer the template, the target query, and the fields available per page. ## Linking and copy in generated sets Large sets fail in ways a single page review won't show: - Related-link orphans: a related block ranked by similarity sends most links to a few popular pages and leaves the long tail with none. Count inbound links per page from the template's real links, and make one slot deterministic, such as the next page in catalog order, so every page has at least one. - Location-filtered hubs: Googlebot's [default IP addresses are in the US and it sends no `Accept-Language`](https://developers.google.com/search/docs/specialty/international/locale-adaptive-pages), so a hub that filters by visitor country shows it only the US subset. Pages outside that subset end up linked only from the sitemap unless a crawl path ignores location, such as related links or an unfiltered index. - Parent pages: a summary written for one child, such as an episode or a product variant, misdescribes its parent. Write the parent's description from fields aggregated across its children. - Spam policy names: Google's [spam policies](https://developers.google.com/search/docs/essentials/spam-policies) call the two failures doorway abuse, near-identical pages built for similar queries, and scaled content abuse, many low-value pages made mainly to rank however they were produced. If an LLM writes intros, titles, descriptions, or alt text, Google asks you to [fact check it before publishing](https://developers.google.com/search/docs/fundamentals/using-gen-ai-content), so review a sample from every pattern. - Hosted sets: a third-party set published on an established domain mainly for that host's ranking signals falls under Google's site reputation policy, and moving a demoted set to a new subdomain or folder counts as circumvention. - Empty and list fields: when a list is empty, branch the sentence instead of printing a dangling colon, and join lists with `Intl.ListFormat` in the page's locale so they read natively in each language. ## Sources - Google Search Central, Spam policies: https://developers.google.com/search/docs/essentials/spam-policies - Google Search Central, Creating helpful, reliable, people-first content: https://developers.google.com/search/docs/fundamentals/creating-helpful-content - Google Search Central, Sitemaps overview: https://developers.google.com/search/docs/crawling-indexing/sitemaps/overview
© fcakyon · MIT · fcakyon/claude-codex-settings
Similar skills
All skills for SEO and GEO →programmatic-seo by other authors comes first.